One of the most challenging problems in color image processing is adjusting the color gains of a system to compensate for variations in illumination spectra incident on an image sensor. This process is typically known as white balance processing. The human eye and brain are capable of “white balancing.” If a human observer takes a white card and exposes it to different kinds of illumination, it will look white even though the white card is reflecting different colors of the spectrum. If a person takes a white card outside, it looks white to the person's eye. If a person takes a white card inside and views it under fluorescent lights, it still looks white. When viewed under an incandescent light bulb, the card still looks white to the human eye. Moreover, even when placed under a yellow light bulb, within a few minutes, the card will look white. With each of these light sources, the white card is reflecting a different color spectrum, but the human brain is smart enough to know that the card looks white.
Obtaining the same result with a camera or other image-capturing device having an image sensor is much harder. When the white card moves from light source to light source, an image sensor “sees” different colors under the different lights. Consequently, when an image-capturing device, e.g., a digital camera, is moved from outdoors (sunlight) to indoor fluorescent or incandescent light conditions, the color in the image shifts. If the white card looks white when indoors, for example, it might look bluish outside. Alternatively, if the card looks white under fluorescent light, it might look yellowish under an incandescent lamp.
The white balance problem stems from the fact that spectral emission curves of common sources of illumination are significantly different from each other. For example, in accordance with Plank's law, the spectral energy curve of the sun is shifted towards the shorter wavelengths relative to the spectral energy curve of an incandescent light source. Therefore, the sun can be considered to be a “blue-rich” illuminator while an incandescent bulb can be considered to be a “red-rich” illuminator. As a result, if the image processing settings are not adjusted, scenes illuminated by sunlight produce “bluish” imagery, while scenes illuminated by an incandescent source appear “reddish”.
After a digital camera (i.e., an image-capturing device) captures an image, the circuitry within the camera performs image processing to compensate for changes in illumination spectra. To compensate for changes in illumination spectra, the gains of the color channels of, e.g. R,G,B, of image processing systems and/or image sensors are adjusted. This adjustment is usually performed by the image processing systems to preserve the overall luminance (brightness) of the image. As a result of proper adjustment, gray/white areas of the image appear gray/white on the image-capturing device (hence the term “white balance”).
In the absence of specific knowledge of the spectra of the illumination source, this adjustment can be performed based on automatic white balance (AWB) statistics. Automatic white balance statistics are based on a statistical analysis of the pixels in the image itself to obtain information about the luminance of colors in the image. The statistical analysis selects a sample of pixels in the image by applying one or more criteria. The values of the pixels that meet the criteria are then used to obtain the color balance statistical information. The image-capturing device can initiate a white balance operation and perform color correction on the image based on the automatic white balance statistics. That is, the collected statistics are compared to expected values and the results of the comparison are used to correct the white balance in the image.
For obtaining automatic white balance statistics, pixels must be selected. One approach to selecting pixels is a white point estimation method. White point estimation can be determined by applying a known gray world model. The gray world model is premised on having the entire image balancing out to gray, i.e., the average color of the image balances out to gray, where gray comprises equivalent amounts of red, green, and blue components. In applying the gray world model for white point estimation, the white point chromaticity corresponds to the average image chromaticity. Since gray is a neutral tone, any variations from the neutral tone in the illumination spectra would be adjusted accordingly.
Several selecting criteria for the selecting pixels are used to obtain automatic white balance statistics. For white point estimation, one selecting criterion requires that only pixels within a white area of the image are selected for the automatic white balance statistics. One method of applying the selecting criterion includes determining a white area of the image sensor, which can be specified during manufacturing by calibrating white curves within the white area. One approach to calibrate a white curve is to take pictures of GretagMacBeth Color Rendition Chart (or similar chart) at different light sources (i.e., different color temperatures) and plot coordinates for gray patches (i.e., applying the gray world model). Since it is known what the colors are supposed to look like (from the chart), raw color data is determined for the image sensor. By applying the gray world model, the coordinates for the gray patches identify a white area for the image sensor; accordingly, a white curve within the white area can be specified for the image sensor.
FIG. 1a illustrates a Log(B/G) vs. Log(R/G) diagram (where B, G, and R are the colors blue, green, and red, respectively) for each GretagMcBeth color square received for a specific image sensor. A series number corresponds to a GretagMcBeth color square number. B/G and R/G ratios are calculated for each row of pixel data received after applying unity analog and digital gain to each color channel at four different illumination source color temperatures ranging from 2800K to 6500K. Series 19-22 are received from the gray GretagMcBeth chart zones and define the white area of the image sensor. Series 1-18 are received from other colors, e.g., red, blue, and green, of the GretagMcBeth chart zones and define the other color chart zones.
FIG. 1b illustrates a white curve within the white area of an image sensor being defined by four nodes plotted in the two-dimension space, i.e., an x and y coordinate space. Each node is associated with coordinates, e.g., Log(B/G) vs. Log(R/G).
Additionally, a threshold distance, shown in FIG. 1b, is specified for the image sensor to determine the boundary of the white area for the image sensor. The threshold distance can be used to determine if a pixel selected from the image is within the white area of the image. The coordinates defining the white curve and the threshold distance are stored in storage areas. The coordinates and threshold distance can later be retrieved by image processing systems and used, for example, as part of the selecting criterion process for white point estimation.
A white curve can also be defined by other coordinates, such as, B/G vs. R/G; Log 2 (B/G) vs. Log 2 (R/B); Y vs. X; (R−B)/Y vs. (R+B−2G)/Y. Each pixel considered for automatic white balance statistics has to be tested to determine if it is within the white area of the image.
A method and system for selecting pixels for automatic white balancing that do not require a large amount of processing resources for selecting are desirable.